Sharp Transformation across Morphotropic Phase Boundary in Sub‐6 nm Wake‐Up‐Free Ferroelectric Films by Atomic Layer Technology

Abstract Atomic layer engineering is investigated to tailor the morphotropic phase boundary (MPB) between antiferroelectric, ferroelectric, and paraelectric phases. By increasing the HfO2 seeding layer with only 2 monolayers, the overlying ZrO2 layer experiences the dramatic phase transition across the MPB. Conspicuous ferroelectric properties including record‐high remanent polarization (2Pr ≈ 60 µC cm−2), wake‐up‐free operation, and high compatibility with advanced semiconductor technology nodes, are achieved in the sub‐6 nm thin film. The prominent antiferroelectric to ferroelectric phase transformation is ascribed to the in‐plane tensile stress introduced into ZrO2 by the HfO2 seeding layer. Based on the high‐resolution and high‐contrast images of surface grains extracted precisely by helium ion microscopy, the evolution of the MPB between tetragonal, orthorhombic, and monoclinic phases with grain size is demonstrated for the first time. The result indicates that a decrease in the average grain size drives the crystallization from the tetragonal to polar orthorhombic phases.

Section 1: FE characteristics of the H 8 Z, H 10 Z, and H 14 Z devices with higher applied voltages to ensure full switching of polarization Higher voltages have been applied to measure the P-V, J-V, and ε r -V curves of the H 8 Z, H 10 Z, and H 14 Z samples to ensure saturated polarization, as shown in Figure S1.Please be aware that the maximum applied voltage in each curve was chosen to keep an identical maximum electric field of 4.7 MV/cm, and so the maximum voltage drops across the ZrO 2 layers are the same in all the samples because the ZrO layers are equivalent in thickness.It can be seen from Figure S1 that the hysteresis and switching current loops of each sample are highly overlapped with almost the same remanent polarizations and coercive voltages, indicating that the applied voltage of ±2.8 V is sufficient for all the samples to reach fully switched polarization in Figure 2. Figure S1 also shows that the dielectric constants (ε r ) of each sample extracted at the maximum voltages remain nearly identical even if higher voltages were applied in the H 8 Z, H 10 Z, and H 14 Z samples.

Figure S1 |
The pristine-state P-V, J-V, and ε r -V curves of the H 8 Z, H 10 Z, and H 14 Z devices.Higher voltages were applied to ensure that the polarization is fully switched.Notice that the maximum applied voltage (3.0 V for H 8 Z, 3.1 V for H 10 Z, and 3.4 V for H 14 Z) in each curve was selected to maintain the same maximum electric field of 4.7 MV/cm.As a result, the maximum voltage drops across the ZrO layers are identical in all samples due to the equal thickness of the ZrO 2 layers.Section 2: The polarity asymmetry of the P-V curves It has been reported that the different work functions of the top and bottom electrodes in MIM devices induce a built-in electric field, which results in the polarity asymmetry of the P-V curves [S1] .In this study, the oxygen species during the deposition of the HZ layer by plasma-enhanced ALD and the oxygen diffusion during the post-deposition annealing would result in the formation of the RuO 2 interfacial layer between the Ru bottom electrode and the HZ layer [S2] .As reported in the literature, the work function of RuO 2 (~5.1 eV) is higher than that of the Ru top electrode (~4.6 eV) [S3] .As a result, the difference between the work functions of RuO 2 on the bottom electrode and the Ru top electrode gives rise to the obvious shift of the P-V curves toward the positive voltage direction in the H 6 Z, H 8 Z, H 10 Z, and H 14 Z samples [S2b] .the oxygen-deficient region, which is thereby correlated with V o [S6] .The surface chemisorbed oxygen such as hydroxyl (−OH) and carbonyl (−CO) [S4b, S7] is responsible for the O III with the highest binding energy.
However, the O III signal is absent in our case due to the removal of the surface chemisorbed oxygen by the in-situ sputtering before the XPS measurement.
It can be seen from Figure S3 that the incorporation of the HfO 2 seeding layer leads to a significant reduction in oxygen vacancies, and the amount of oxygen vacancies decreases with the HfO 2 thickness.It is worth noting that even just two ALD cycles of HfO 2 can substantially decrease the oxygen vacancies from 12.6% in H 0 Z to 5.4% in H 2 Z.The HfO 2 seeding layer may suppress the scavenging of oxygen by the bottom electrode, which could be the reason for the decrease in the content of oxygen vacancies in overlying ZrO 2 layer.
Both the first-principle calculation and experimental studies have addressed that the crystalline structures in the HfO 2 -and ZrO 2 -based thin films can be actually affected by the amount of oxygen vacancies [S8] .Based on the simulation using density functional theory, the formation energies of both the oand t-phases relative to the m-phase decrease with an increase in the concentration of oxygen vacancies [S8a, S8d] .Furthermore, by tuning the ozone dose during the ALD process of depositing HfO 2 -and ZrO 2 -based thin films, the effect of oxygen vacancies can be identified experimentally [S8c-e, S9] .Insufficient ozone dose results in a large amount of oxygen vacancies, thereby substantially facilitating the formation of the t-phase [S8c, S8e] .As the ozone dose increases, there exists an optimal window of the quantity of oxygen vacancies for stabilization of the o-phase [S8c, S8e, S8f, S10] .With a further increase in the ozone dose, the m-phase is present since the lack of oxygen vacancies is favorable for the formation of the m-phase [S8c-e] .As shown in Figure S3, the content of oxygen vacancies decreases steadily with an increase in the thickness of the HfO 2 seeding layer, and the corresponding crystalline structure undergoes the transformation from a mixture of t-and ophases, to predominant o-phase, and ultimately to the coexistence of o-and m-phases, which are highly consistent with the first-principle prediction and experimental results as mentioned above.However, the crystalline phases can be affected not only by oxygen vacancies, but also by other factors such as doping, strain, and surface energy [S10a, S11] .In this study, it has been experimentally demonstrated that the HfO 2 seeding layer has a substantial influence on the phase transition temperature and the strain in the films, both of which also hold considerable significance in the resulting crystalline phase.Section 5: Endurance and wake-up free performance of all the FE devices In order to gain a comprehensive understanding of the relevance between the wake-up effect and the quantity of oxygen vacancies, the endurance was examined on all the FE (H 6 Z, H 8 Z, H 10 Z, and H 14 Z) devices.The endurance test was performed by the cycling bipolar triangular voltage with a magnitude of ±2.8 V at a frequency of 2 kHz.It can be observed from Figure S4 that the 2P r values keep almost identical throughout ~10 5 cycles, indicating the wake-up-free performance of the H 6 Z, H 8 Z, H 10 Z, and H 14 Z devices.
Because the amount of oxygen vacancies (4.5% to H 6 Z, 4.0% to H 8 Z, 3.9% to H 10 Z, and 3.8% to H 14 Z) are significantly suppressed by the HfO 2 seeding layer, the wake-up-free feature is highly correlated with the low contents of oxygen vacancies in the FE layers [S9, S12] .Nevertheless, the 2P r degrades abruptly when the endurance cycles exceed 10 6 for all the devices.Since oxygen vacancies have been proposed to be in favor of the stabilization of the metastable polar o-phase [S8a, S8c, S13] , the fatigue in the HZ thin films should be associated with the low amount of oxygen vacancies in the FE thin films.

Figure S2 |
Figure S2| Out-of-plane, grazing incident, and in-plane XRD patterns of all the samples for the identification of residual stress.The apparent shift from the out-of-plane and grazing incident XRD peaks to the in-plane XRD peak toward the lower diffraction angle can be observed in the H 6 Z sample, indicating the presence of significant in-plane tensile stress.Apart from the H 6 Z sample, the other samples show minor or irregular offsets among the out-of-plane, grazing incident, and in-plane diffraction peaks.

Figure S3 |
Figure S3 | The O 1s XPS spectra and the contents of oxygen vacancies in all of the samples.The proportions of oxygen vacancies are (a) 12.6% in H 0 Z, (b) 5.4% in H 2 Z, (c) 5.1% in H 4 Z, (d) 4.5% in H 6 Z, (e) 4.0% in H 8 Z, (f) 3.9% in H 10 Z, and (g) 3.8% in H 14 Z samples.The HfO 2 seeding layer plays a crucial role in suppressing oxygen vacancies.

Figure S4 |
Figure S4| The endurance and wake-up-free operation of the FE devices.The 2P r values keep nearly constant within 10 5 wake-up cycles, which reveals the wake-up-free ferroelectricity of the H 6 Z, H 8 Z, H 10 Z, and H 14 Z devices.After 10 6 cycles, the 2P r begins to degrade in all the samples.

Figure S5 |
Figure S5 | Low-magnification cross-sectional HRTEM image of the HZ sample.

Section 3: The XRD patterns of the t(101)/o(111) crystallographic planes in the out-of-plane, grazing incident, and in-plane scanning modes
The out-of-plane, grazing incident, and in-plane XRD patterns of all the samples are shown in Figure S2. Upon examination of the out-of-plane, grazing incident, and in-plane XRD patterns in each sample, there are no distinct and regular offsets in the XRD peaks from t(101)/o(111) crystallographic planes except for the H 6 Z sample.The lack of noticeable in-plane tensile stress corresponds to the presence of the predominant t-phase in the H 0 Z, H 2 Z, and H 4 Z samples, as well as the predominant m-phase in the H 10 Z and H 14 Z samples.As a result, it can be deduced that the presence of significant in-plane tensile stress is favorable for the formation of the FE o-phase in the H 6 Z sample.